An experimental study on the self-adaption mechanism used by evolutionary programming

Evolutionary programming (EP) is one of the most important methods for numerical optimization. Its main technique is the combination of mutations and the self-adaption mechanism. In the past years, studies on EP focused on how to improve the efficiency of mutations with different probability distributions, and few of them touched on the question that the self-adaption mechanism did not work sometimes, but simply followed the original suggestion. So far, no experimental results have shown why this is a question. This paper firstly gives a primary analysis on the behavior of the self-adaption mechanism, and then presents experimental evidences to show why its adaptive ability is doubtful.